Sentiment-Analysis-API-with-HuggingFace

Containerized sentiment analysis API using HuggingFace’s DistilBERT model, powered by FastAPI and Uvicorn. Includes parallel request demonstration via Jupyter Notebook.

1. Project Overview This project deploys a sentiment analysis model using Hugging Face's transformers library, served via a FastAPI application. The system is containerized with Docker and supports concurrent inference requests. A Jupyter Notebook demonstrates parallel POST requests to the endpoint.

2. Model Used distilbert-base-uncased-finetuned-sst-2-english
   https://huggingface.co/distilbert/distilbert-base-uncased-finetuned-sst-2-english#how-to-get-started-with-the-model

3. Tech Stack
   • Python 3.10+
   • FastAPI
   • HuggingFace Transformers
   • Docker
   • Uvicorn (ASGI server)
   • Jupyter Notebook
   • aiohttp and asyncio (for parallel requests demo)
   

4. Project Structure
   main.py # FastAPI app with /predict endpoint
   Dockerfile # Instructions for containerizing the app
   requirements.txt # Python dependencies
   .dockerignore # Files to exclude from Docker builds
   parallel_requests_demo.ipynb # Notebook to test multiple requests
   README.md # Project documentation
   

5. How to Run the Server
   Using Docker:
   docker build -t sentiment-api .
   docker run -p 8000:8000 sentiment-api
   Access the API at http://localhost:8000/docs.
   

6. API Usage
   Endpoint: POST /predict
   Payload example:
   {
   "text": "This product is amazing!"
   }
   Response example:
   {
   "label": "POSITIVE",
   "score": 0.9993
   }
   

7. Parallel Request Demo The parallel_requests_demo.ipynb file shows how to simulate multiple users hitting the endpoint in parallel using aiohttp and asyncio.

8. Future Enhancements
   • Add support for batch inference.
   • Deploy on cloud (AWS/GCP) with autoscaling.
   • Integrate logging and monitoring tools.